INDEX

Explanations

terms related to specific names or titles, particularly those including the word "Box"

oai_token-act-pair · gpt-3.5-turbo

references to boxes and their associated numbers

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ittee

-0.73

merce

-0.72

FUL

-0.72

ufact

-0.70

puter

-0.70

asury

-0.68

ictional

-0.67

opoulos

-0.64

nee

-0.64

 ditch

-0.63

POSITIVE LOGITS

Box

1.06

er

1.05

es

1.04

Box

1.02

boxes

1.00

box

1.00

esy

0.97

wra

0.96

cars

0.96

sets

0.95

Activations Density 0.015%

terms related to specific names or titles, particularly those including the word "Box"

references to boxes and their associated numbers

No Comments

No Known Activations